Íslenskt mál og almenn málfræði - 01.01.1984, Side 143

141
Model ofModern Icelandic Syllable Types
nucleus and coda in the way described above, I will now proceed
to make an attempt to give a classification of the nuclei with the aid
of the parameters mentioned at the outset.
The first parameter, that of voicing, will be assumed to have two
values, what may be called held and checked. (These terms are remin-
iscent (and derivative) of terminology used by Coates (1979, 1980),
but used here as referring to a more ‘abstract’ level of phonological
organization. I am here dealing with traditional ‘structure’ rather than
actual behaviour.)
The value held for a nucleus means that the nucleus is voiced
throughout, as in vor [vo:r] ‘spring’ te [the:] ‘tea’, vald [val$] ‘power’,
whereas voicing is said to be checked in forms like vors [vojs] ‘spring’
(gen.), mark [majg] ‘goal’, epli [ehþll] ‘apple’, hatt [hah^] ‘hat’, hest
[hes^] ‘horse’.
This dichotomy of stressed nuclei I will, for the purposes of this
paper, assume to be the most basic, or highest in the hierarchy, and
this is why this (two-place) system is the leftmost in Diagram 1. The
classification obtained by this dichotomy seems to be relatively agree-
able intuitively. In particular, it forms a neat ‘natural class’ out of
the stressed syllables that have a voiceless interval before a stop or
fricative. The commonness of the appearance of voicelessness at this
place in the articulation is seen as a basic characteristic of Icelandic
phonology.2
2 This dichotomy between held (fully voiced) and checked (devoiced) nuclei will
be assumed to be enough for the purposes of this paper, although it seems likely that
from a more phonetic point of view this dichotomy wiH have to be revised and replaced
by a trichotomy involving a ternary feature of glottal activity, having values something
like ‘glottal stop’ — ‘fully voiced’ and ‘voiceless’ (cf. Ladefoged 1971:7-21, 92). The
third category ‘glottal stop’ would be appropriate for forms containing stops immediately
following the vowel, as in Bjarni [bjadni] ‘a man’s name’, lögga [lceg:a] ‘policeman’.
These stops are regularly glottalized, and sometimes they lose their oral articulation
when nasals follow. Thus we get pronunciations like [bjaTni] for Bjarni [bjadni]. Al-
though it may seem strange to allow this contradiction in the analysis to go unsettled,
it is not as harmful as it may seem, since although articulatorily quite different, glottaliz-
ation and voicelessness with an open glottis can be (and often are) classified together
under the rubric voiceless, which from the acoustic point of view may have very similar
effects. Also, glottalization can be analysed as a feature of the segments in question,
although this would be somewhat contrary to the (half-hearted) spirit of this paper
(cf. p. 148-9). Part of the reason why this question is left unsettled here is the fact
that the phonetics of these penomena need closer investigation.